القائمة
Review code for patterns known to cause flaky tests. Use when the user asks to review tests or a change for flakiness, or when reviewing a diff that adds or modifies tests.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Generate CockroachDB SQL and migrations for schema changes. Use when creating migrations, updating the database schema, or when the user mentions migrations, schema changes, or dbinit.sql.
Add a new Nexus background task. Use when the user wants to create a periodic background task in Nexus that runs on a timer.
Add a new version to a Dropshot HTTP API. Use when the user wants to make versioned changes to a Dropshot-managed HTTP API, such as adding, making changes to, or removing an API endpoint.
| name | flake-review |
| description | Review code for patterns known to cause flaky tests. Use when the user asks to review tests or a change for flakiness, or when reviewing a diff that adds or modifies tests. |
Review code for patterns known to cause flaky tests in Omicron.
Read docs/flake-patterns.adoc in full. It describes each known flake pattern: what it is, why it is bad, how to fix it, and example commits. Do not work from memory; the catalog is the source of truth and may have gained patterns since your last memory or training data.
If the user pointed at specific files, tests, or a PR, review those. Otherwise, prompt the user to ascertain the scope of the diff. This can be:
Uncommitted changes: Changes not yet committed.
git diff (unstaged) or git diff --cached (staged)jj diff --gitThis commit only (stacked diff workflow): Changes are in the current commit only.
git diff HEAD^jj diff --git --from @--This branch (feature branch workflow): Changes span the entire branch.
git diff $(git merge-base HEAD main)jj diff --git --from 'fork_point(trunk() | @)'Focus on test code and on test-support code (test utilities, simulated components, test configs), but remember that some flakes are genuine bugs in the system under test, so also consider production code that the changed tests exercise.
For each pattern in the catalog, check whether the code under review exhibits it. Remember that you can examine the example commits by running git show <commit-hash> or jj show --git <commit-hash>.
For each finding, report:
file:line).wait_for_condition, wait_for_watch_channel_condition, now_db_precision()).If a finding looks like a bug in the system under test rather than the test itself, say so explicitly.